Function Computation via Over-the-Air Computation: Practical System Design

Abstract: Future wireless networks will need to contain millions of nodes, machines, and devices and provide high-speed and low-latency communication services. The traditional wireless network architecture design separates the communication and computation process, which has low utilization of wireless resources. In recent years, a new technology, Over-the-Air Computation(OAC), has taken advantage of the superposition properties of wireless channels and can perform the computation of specific functions without aggregating independent data, thus greatly reducing the time required to operate network functions. Many theoretical studies have found two ways to implement Over-the-Air computation: analog OAC and digital OAC. This thesis addresses the theoretical study of analog OAC by building an experimental platform and verifying the theory with the Software-defined Radio(SDR) device ADALM-PLUTO. This thesis provides an idea for building the experimental platform and a sample-level simulator to analyze the effect of the number of devices and the coding method on the transmission results. The paper then describes the process of conducting hardware experiments on the experimental platform and verifies the conclusions drawn by the simulator by analyzing the mean square error (MSE) of multi-times transmission results. This thesis also describes and analyzes the problems that arise during the construction of the hardware experimental platform and proposes alternative solutions. In addition, the thesis discusses the experimental platform's limitations and possible future improvement methods.